chapter 4 numerical differentiation and integration 1/16 given x 0, approximate f ’(x 0 ). h xfhxf...

16
Chapter 4 Numerical Differentiation and Integration 1/1 6 Given x 0 , approximate f (x 0 ). h x f h x f x f h ) ( ) ( lim ) ( ' 0 0 0 0 h x f h x f x f ) ( ) ( ) ( ' 0 0 0 x 0 x 1 h x 1 x 0 h h h x f x f x f ) ( ) ( ) ( ' 0 0 0 4.1 Numerical Differen tiation forward backwar d

Upload: eileen-tucker

Post on 26-Dec-2015

254 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration

1/16

Given x0, approximate f ’(x0).

h

xfhxfxf

h

)()(lim)(' 00

00

h

xfhxfxf

)()()(' 00

0

x0 x1

h

x1 x0

hh

hxfxfxf

)()()(' 00

0

4.1 Numerical Differentiation

forward

backward

Page 2: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Numerical Differentiation

Approximate f(x) by its Lagrange polynomial with interpolating points x0 and x0 + h.

f (x) =

)(2

))((

))(())((

00

00

00

00

00

xfhxxxx

xhx

xxhxf

hxx

hxxxf

f ’(x) =

)(2

))((

)(2

)(2)()(

00

000

x

x

fdx

dhxxxx

fhxx

h

xfhxf

f ’(x0) = )(2

)()( 00 fh

h

xfhxf O(h)

2/16

Page 3: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Numerical Differentiation

Approximate f(x) by its Lagrange polynomial with interpolating points { x0, x1, …, xn }.

)()!1(

))...(()()()( )1(0

0x

nnn

kkk f

n

xxxxxLxfxf

f ’(xj) =

n

jkk

kjj

nn

kjkk xx

n

fxLxf

0

)1(

0

)()!1(

)()()(

Note: Note: In general, more evaluation points produce greater In general, more evaluation points produce greater accuracy. accuracy. On the other hand, the number of functional evaluations On the other hand, the number of functional evaluations grows and the roundoff error increases. Hence the numerical grows and the roundoff error increases. Hence the numerical differentiation is differentiation is unstableunstable!!

3/16

Page 4: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Numerical Differentiation

Example: Given three points x0, x0 + h, and x0 + 2h, please derive the three-point formulae for each of the points.

)(3

)2(2

1)(2)(

2

31)( 0

)3(2

0000 fh

hxfhxfxfh

xf

)(6

)2(2

1)(

2

11)( 1

)3(2

000 fh

hxfxfh

hxf

)(3

)2(2

3)(2)(

2

11)2( 2

)3(2

0000 fh

hxfhxfxfh

hxf

Symmetric to formula 1, with h < 0.

)(6

)()(2

1)( 1

)3(2

000 fh

hxfhxfh

xf

x–1 x1x0Five-point formulae are given on p.171

4/16

Page 5: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Numerical Differentiation

Approximate f ”(x0)

Consider Taylor expansions of f(x0 + h) and f(x0 – h) at x0:

41

)4(30

20000 )(

24

1)(

6

1)(

2

1)()()( hfhxfhxfhxfxfhxf

41

)4(30

20000 )(

24

1)(

6

1)(

2

1)()()( hfhxfhxfhxfxfhxf

)(12

)()(2)(1

)( )4(2

00020 fh

hxfxfhxfh

xf

HW: p.176-177 #7, 13

Excuses for not doing homework

I could only get arbitrarily close to my textbook.

I couldn't actually reach it.

5/16

Page 6: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Elements of Numerical Integration

4.3 Elements of Numerical Integration

Approximate b

adxxfI )( -- Numerical Quadrature

Integrate the Lagrange interpolating polynomial of f (x) instead.

Idea

Select a set of distinct nodes a x0 < x1 <…< xn b from [a, b]. T

he Lagrange polynomial is

n

kkkn xLxfxP

0

)()()(

b

a

b

a k

n

kk dxxLxfdxxf )()()(

0

Ak

b

a kj xx

xx

k dxAjk

j

)(

)(

b

a

n

kk

xn

b

a n

b

a n

b

a

n

kkk

dxxxn

f

dxxRdxxPxf

xfAdxxf

fR

0

)1(

0

)()!1(

)(

)()]()([

)()(

][

Error

interpolatory quadrature

6/16

Page 7: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Elements of Numerical Integration

Definition: The degree of accuracy, or precision, of a quadrature formula is the largest positive integer n such that the formula is exact for xk for each k = 0, 1, …, n.

Example: Consider the linear interpolation on [a, b], we have

)()()(1 bfab

axaf

ba

bxxP

221

abAA

)]()([

2)( bfaf

abdxxf

b

a

f(x)

a b

f(a)f(b)trapezoidal rule

Please determine the precision of this formula.

Solution: Consider xk for each k = 0, 1, …

x0 = 1: b

aabdx1 ]11[2 ab

=

x : =

x2

2

22 abb

adxx ][2 baab

32 33 ab

b

adxx ][ 22

2 baab Degree of Precision = 1

7/16

Page 8: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Elements of Numerical Integration

For equally spaced nodes: nin

abhhiaxi ,...,1,0,,

dxxx

xxA

nx

xij ji

ji

0 )(

)(

n

ji

inn

ji

dtjtinin

abdth

hjihjt

00)(

)!(!)1)((

)()(

Let htax

Cotes coefficient )(niCNote: Cotes coefficients does not depend on either Note: Cotes coefficients does not depend on either ff((xx) or [) or [aa, , bb], ],

and can be determined by and can be determined by nn and and ii only. Hence we can find these only. Hence we can find these coefficients from a table. The formulae are called coefficients from a table. The formulae are called Newton-Cotes Newton-Cotes formulaeformulae..

21

,21 )1(

1)1(

0 CCn = 1: )]()([2

)( bfafab

dxxfb

a

Trapezoidal Ruledxbxaxf

fRb

a

x ))((!2

)(][

1,],[,)(

121 3 ab

hbafh

Precision = 1

8/16

Page 9: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Elements of Numerical Integration

n = 2:61

,32

,61 )2(

2)2(

1)2(

0 CCC

)]()(4)([6

)( 2 bffafab

dxxf bab

a

Simpson’s Rule

Precision = 3

2,),(,)(

901

][ )4(5 abhbafhfR

n = 3: Simpson’s 3/8-Rule. Precision = 3, and )(803

][ )5(5 fhfR

n = 4: Cotes Rule. Precision = 5, and )(945

8][ )6(7 fhfR

Theorem: For the (n+1)-point closed Newton-Cotes formula, there exists (a, b) for which

nnnn

kkk

b

adtnttt

n

fhxfAdxxf

0

2)2(3

0

))...(1()!2(

)()()(

if n is even and f Cn+2[a, b], and

nnnn

kkk

b

adtnttt

n

fhxfAdxxf

0

)1(2

0

))...(1()!1(

)()()(

if n is odd and f Cn+1[a, b].

HW: p.195 #7, 9, 11, 13

9/16

Page 10: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Composite Numerical Integration

4.4 Composite Numerical Integration

Due to the oscillatory nature of high-degree polynomials, piecewise interpolation is applied to approximate f(x) a piecewise approach that uses the low-order Newton-Cotes formulae.

Haven’t we had enough formulae? What’s up

now?

Oh come on, you don’t seriously consider

h=(ba)/2 acceptable, do you?

Why can’t you simply refine the partition if you have to

be so picky?

Don’t you forget the oscillatory nature of high-

degree polynomials!Uh-oh

Composite Trapezoidal Rule: ),...,0(, nkhkaxn

abh k

Apply Trapezoidal Rule on each [xk – 1, xk]:

nkxfxfxx

dxxf kk

x

x

kkk

k

,...,1,)]()([2

)( 11

1

1

1

)()(2)(2

n

kk bfxfaf

h

b

a

n

kkk xfxf

hdxxf

11 )]()([

2)( = Tn

),(),()(12

)()(

12)](

12[][

2

12

1

3

bafabh

n

fab

hf

hfR

n

kkn

kk

/*MVT*/

10/16

Page 11: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Composite Numerical Integration

Composite Simpson’s Rule: ),...,0(, nkhkaxn

abh k

)]()(4)([6

)( 121

1

kkk

x

xxfxfxf

hdxxf

k

k

kx21kx 1kx 4

44

44

])()(2)(4)([6

)(1

0

2

01

21

n

k

n

kkk

b

abfxfxfaf

hdxxf = Sn

)(2180

][ )4(4

fhab

fR

Note: Note: To simplify the notation, we may let To simplify the notation, we may let nn’ = 2’ = 2nn. Then. Then

andandhkaxh

n

abh k

,2

])()(2)(4)([3

kodd keven

kkn bfxfxfafh

S

11/16

Page 12: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Composite Numerical Integration

Composite integration techniques are all stable.

Example: Consider the Simpson’s Rule with n subintervals on [a, b]. Assume that f (xi) is approximated by f *(xi) such that f (xi) = f *(xi) + i for each i = 0, 1, …, n. Then the accumulated error e(h) is

]24[3

)( 0 kodd keven

nkk

hhe

If | i | < for all i = 0, 1, …, n, then

)(])12/(2)2/(4[3

)( abnhnnh

he

When we refine the partition to ensure accuracy, the increased computation will NOT increase the roundoff error.

12/16

Page 13: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Composite Numerical Integration

Example: Use Trapezoidal rule and Simpson’s rule with n = 8 to

approximate dxx

1

0 21

4

)1()(2)0(16

1 7

18 fxffT

kk

8

kxk where

= 3.138988494

)1()(2)(4)0(

24

1

odd even4 fxfxffS kk

8

kxk where

= 3.141592502

Solution:

When programming we usually keep dividing the subintervals into 2 equally spaced smaller subintervals. That is, take n = 2k for k = 0, 1, …

When k = 9, T512 = 3.14159202

48 3

1

3

4TT = 3.141592502 = S4

HW: p.204 #7(a)

(b)

13/16

Page 14: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Romberg Integration

4.5 Romberg Integration

4

12

n

n

TI

TI

][4

1)()(

12

1

2][

2

2 fRfabh

fR nn

),()(12

][2

fabh

fRn Since the error of Trapezoidal rule is

when we reduce the length of each subinterval into a half,

nnnn TT

TTI

3

1

3

4

14

42

2

Solve for I : = Sn

In general: nnn S

TT

14

4 2n

nn CSS

14

422

2

nnn R

CC

14

432

3

Romberg sequence

Romberg method:

< ?

< ?

< ?

… … … … … …

T1 =)0(

0T

T8 =)3(

0T

T4 =)2(

0T

T2 =)1(

0T S1 =)0(

1T

R1 =)0(

3T

S2 =)1(

1T C1 =)0(

2T

C2 =)1(

2T S4 =)2(

1T

14/16

Page 15: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Lab 08. Shape Roof

Time Limit: 2 seconds; Points: 4

The kind of roof shown in Figure 1 is shaped from plain flat rectangular plastic board in Figure 2.

Figure 1 Figure 2

The transection of the roof is a sine curve with altitude l centimeters. Given the length of the roof, your task is to calculate the length of the flat board needed to shape the roof.

15/16

Page 16: Chapter 4 Numerical Differentiation and Integration 1/16 Given x 0, approximate f ’(x 0 ). h xfhxf xf h )()( lim)(' 00 0 0    x0x0 x1x1 h x1x1 x0x0

Chapter 4 Numerical Differentiation and Integration -- Richardson’s Extrapolation

4.2 Richardson’s Extrapolation

Generate high-accuracy results while using low-order formulae

Suppose that for some h 0, we have a formula T0(h) that approximates an unknown I, and that the truncation error has the form:

T0(h) I = 1 h + 2 h2 + 3 h3 + …Replace h by half its value, we have

T0(h/2) I = 1 (h/2) + 2 (h/2)2 + 3 (h/2)3 + …

Q: How to improve the accuracy from O(h) to O(h2) ?

...4

3

2

1

12

)()(2 33

22

020

hhIhTT h

...12

)()(2)( 3

22

1020

1

hhIhTT

hTh

...)( 42

312 hhIhT

12

)()(22

1212

hTT h

...)( 22

11 mm

m hhIhT 12

)()(2 121

mm

hm

m hTT

16/16